The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection because it processes data from external Figma designs to generate component code.
Ingestion points: Design context is retrieved from user-supplied Figma URLs using the get_design_context tool via the plugin-figma-figma MCP server, as described in references/figma-design-rules.md.
Boundary markers: The SKILL.md workflow includes a mandatory confirmation step (Step 2) where the agent must echo back the understood dialog specification for user approval before any files are created.
Capability inventory: The skill is capable of writing multiple file types (XML, HTL, Java, CSS, and JS) to the project's local directory structure as defined in SKILL.md.
Sanitization: The instructions in references/figma-design-rules.md treat Figma output strictly as a reference, requiring the agent to translate JSX structures into semantic AEM HTL rather than directly executing or copying external code.
[SAFE]: The skill implements a robust 'Configuration Gate Check' that prevents the agent from operating until a local .aem-skills-config.yaml file is verified, ensuring that file operations occur within the intended project scope and paths.
[SAFE]: The reference documentation included with the skill (e.g., references/htl-patterns.md and references/sling-servlet-standards.md) enforces security best practices for all generated code, specifically mandating XSS expression contexts and CSRF token handling.

create-component